The final session of the Capita FutureGov summit began with a rhetorical question posed in a recent speech by Admiral Sir Tony Radakin, outgoing Chief of the Defence Staff: “We tend to think all the risk is with the new and with innovation, but we don't think about the risk of just doing things in the same way.”
This session explored examples of where AI was already delivering value for government departments, with three speakers representing London transport, army recruitment, and internal audit.
Making drivers’ lives easier
The first speaker described the motivation for exploring AI in transport as dating back to 2009, when it delivered value in optimising timetable schedules.
This time round, the focus was on passenger experience and the complexities around toll roads. There are numerous discounts for cleaner vehicles and local residents, but for each, drivers have to prove eligibility. It took five days to process an application while paying the full toll, which was later refunded. Now, image recognition with GenAI scans a registration plate and delivers an eligibility decision in five minutes – which delivers significant value for road users.
Accelerating medical checks in recruitment
The second speaker described a transformed recruitment process.
Army personnel need to be able to operate for weeks and months in isolated, austere, filthy environments, under daily physical and mental strain, so people need to be medically fit.
The historical process for screening medical conditions involved funnelling 80,000 applications down to 9,000 by asking for medical records, checking them against standards, and inviting those that meet the standards to an assessment centre for a face-to-face medical examination, and physical, cognitive, and teamwork tests.
It used to take up to 30 days per applicant to get consent to see a medical record – now, there is a digital consent form with a tick box for candidates, which automatically contacts their GP immediately. With the electronic record received, an AI-based tool scans the records and produces a two-page summary for nurses to look at, with highlighted points for in-depth analysis.
This saves 28,000kg of paper every year – the weight of four African elephants. “But more importantly, it gets good people into the Army faster,” said the speaker. Consent happens in minutes, and the records sourcing process has been cut from 30 days to 8 days, saving £1.1m in productive time every year. It also enables safer, better clinical decisions.
Streamlining internal audit
The third speaker is using AI in internal audit across 15 out of 17 government departments.
His team produces around 1500 reports a year, and the AI programme started with the problem of manually reading and producing insights at such a scale. So the team developed AI to analyse the content of, for example, public consultations, and, within an hour, understand insights across hundreds of documents.
The team has also created a risk engine, which is used widely. Using GenAI can help people think in a divergent way about risks, so they focus on the right area. The tool also writes up a risk report and suggests recommendations, and has reduced the time taken from 20 days to four days. Just this one tool has reduced operational costs by 2% - a very substantial return.
The tool encourages people to think about objectives and how to avoid potential consequences – “People become very enthusiastic because they’ve been part of the process from the beginning,” said the speaker.
Unexpected benefits and challenges
The speakers were asked to give examples of unexpected benefits and challenges in their AI deployments.
One speaker reported that there was real enthusiasm to experiment as it opened people’s minds and also yielded bigger results than expected.
Another had also seen how similar use cases can be transferred quickly when other departments hear about the benefits being delivered.
The third speaker relayed that people had become creative with the tools, using them in ways that hadn’t been anticipated. There had also been some initial uncertainty among employees around using AI tools, but this was quickly overcome when some of those who had been sceptical were encouraged to share their positive journeys.
Final advice
The speakers shared some closing thoughts for delegates on implementing AI.
- Being business-led has been incredibly productive – find the use case and get results at low expense. Think carefully about what problem you’re trying to solve and find the right tools.
- Start small and aggregate marginal gains – the value AI can deliver has broken boundaries.
- Focus on the immediate risks – don’t anticipate everything that could possibly go wrong with AI. Ask, what is it you’re worried about? Then choose a specific use case, so the risk assessment applies to just a small subset of activity, then work around it – that’s your starting point.
You can read the first article in this series here: Understanding the AI opportunity in government
You can read the second article in this series here: AI in government: From vision to implementation
Ready to explore how AI can unlock value for your organisation? Capita is hosting a series of free-to-attend AI and Copilot workshops designed to spark open, honest conversations about your hopes, challenges, and how we can help. Whether you're curious about our new AI Catalyst Stack (see Capita.com) or interested in a tailored Copilot workshop, we’d love to hear from you. To register, email bettergovernment@capita.com.